Functional significance of rare neuroligin 1 variants found in autism

Abstract

Genetic mutations contribute to the etiology of autism spectrum disorder (ASD), a common, heterogeneous neurodevelopmental disorder characterized by impairments in social interaction, communication, and repetitive and restricted patterns of behavior. Since neuroligin3 (NLGN3), a cell adhesion molecule at the neuronal synapse, was first identified as a risk gene for ASD, several additional variants in NLGN3 and NLGN4 were found in ASD patients. Moreover, synaptopathies are now known to cause several neuropsychiatric disorders including ASD. In humans, NLGNs consist of five family members, and neuroligin1 (NLGN1) is a major component forming a complex on excitatory glutamatergic synapses. However, the significance of NLGN1 in neuropsychiatric disorders remains unknown. Here, we systematically examine five missense variants of NLGN1 that were detected in ASD patients, and show molecular and cellular alterations caused by these variants. We show that a novel NLGN1 Pro89Leu (P89L) missense variant found in two ASD siblings leads to changes in cellular localization, protein degradation, and to the impairment of spine formation. Furthermore, we generated the knock-in P89L mice, and we show that the P89L heterozygote mice display abnormal social behavior, a core feature of ASD. These results, for the first time, implicate rare variants in NLGN1 as functionally significant and support that the NLGN synaptic pathway is of importance in the etiology of neuropsychiatric disorders.

Fig 1. Identification of P89L substitution in NLGN1.

(A) Pedigree of the family harboring the NLGN1 c.266C>T (p.Pro89Leu) substitution. (B) Sequence electropherograms of family AU0729. Father (1), Mother (2), unaffected sibling (3), affected siblings (4 and 5). Arrows indicate the location of the NLGN1 c.266C>T substitution.

Fig 2. In silico investigation of NLGN1 variants observed in patients with ASD.

(A) Ribbon diagram of the extracellular part of the mouse NLGN1 dimer viewed from the side, based on Protein Data Bank (PDB) entry 3B3Q. The dotted line shows the NLGN1 region for which crystal structure is unavailable. Amino-acid numberings of mouse NLGN1 (human NLGN1) are indicated. The variants assessed in this study are shown in red. The gray box indicates the location of a proline-rich loop of NLGN1. (B) Ribbon diagram of the NLGN1-NRXN1β complex, viewed from the post-synaptic membrane with variants. (C) Enlarged image of the proline-rich loop structure of NLGN1 in which P89 is located (shown in light blue). Previously identified NLGN4 missense variants (G84R, R87W, and G99S) found in ASD are also indicated. (D) The protein alignment of the NLGN1-4 family. The highly conserved P89 residue, mutated in ASD patients, is boxed in red, and pathogenic NLGN4 variants, localized in the same loop, are shown as blue arrows.

Fig 3. Pathogenic NLGN1 variants exhibit abnormal sub-localization and expression.

(A) Fluorescence images of COS7 cells transfected WT or mutant NLGN1 with HA-tag. Endoplasmic reticulum (ER) was stained with calnexin. Three pathogenic variants (P89L, L269P, and G288E) were trapped in ER, and failed to traffic the plasma membrane. Scale bar indicates 10 μm. (B) Representative images of western blots of cell lysates from COS7 cells transfected with HA-tagged NLGN1. NLGN1 was detected by anti-HA tag. The expected molecular weight for the NLGN1: glycosylated mature NLGN1 (~110 kDa), non-glycosylated immature NLGN1 (~100 kDa). (C) Quantitative analysis of the western blots for total (left) and glycosylated (right) NLGN1. The expression of NLGN1 variants is normalized to the corresponding β-actin. NLGN1 variants observed in ASD subjects showed decreased expression level compared to WT. Data represents mean ± S.E.M. of four samples from three independent experiments (one-way ANOVA followed by Tukey-Kramer’s multiple comparisons test, *p<0.05, **p<0.01 compared with WT). Gray, red, and blue bars indicate the non-pathogenic variant, “high-risk”, and “low-risk” pathogenic variants, respectively. (D) Representative image of western blots of conditioned media (collected at 18 and 32 hr after transfection), cell lysates (collected at 32 hr after transfection), and β-actin as an internal control. Both cleaved NLGN1 in media and full-length NLGN1 in cell lysate were detected by anti-HA tag. (E) Quantification of cleaved NLGN1 with the HA-tag inserted in extracellular domain in conditioned media. Relative expression levels at 18 hr post-transfection (left) and 32 hr post-transfection (right) are shown. Cleaved NLGN1 was decreased in three pathogenic variants (P89L, L269P, and G288E) and increased in H786Y variant compared to WT. Data represents mean ± S.E.M. of six samples (one-way ANOVA followed by Tukey-Kramer’s multiple comparisons test, *p<0.05, **p<0.01 compared with WT).

Fig 4. Impaired spine induction by NLGN1 variants observed in patients with ASD.

(A) Representative fluorescence images of hippocampal neurons (DIV14) co-transfected with negative control, WT, and non-pathogenic NLGN1 or pathogenic NLGN1 variants with GFP expression vector. NLGN1 expression of pathogenic variants (P89L, L269P, G288E, and H786Y) was decreased compared to others. (B) Representative images of spines from neurons transfected with negative control, WT, and non-pathogenic NLGN1 or pathogenic variants of NLGN1. (C) Quantification of the number of spines. Spine number in dendrite was significantly lower in four pathogenic variants (P89L, L269P, G288E, and H786Y) compared to WT, and no significant differences between GFP-transfected control and these four variants were observed. Data represent mean ± S.E.M. (*p<0.05, **p<0.01 Tukey-Kramer’s multiple comparisons test). More than 14 neurons are counted for each NLGN1 variant. Scale bar indicates 20 μm for (A) and 5 μm for (B).

Fig 5. Generation of Nlgn1 P89L KI mice.

(A) Schematic of the mouse genomic locus of Nlgn1 showing the target site of Cas9. sgRNA sequence is underlined, and replaced bases in knock-in mice are capitalized. (B) DNA sequence electropherograms of WT and knock-in heterozygote mouse. Red arrow indicates the amino acid substitution from proline to leucine (CCA to CTA of residue 89), and blue arrow indicates a silent mutation for genotyping using restriction enzyme BsrBI. (C-D) Western blots of cortex, hippocampus, and cortical synaptosomal fractions from wild-type, Nlgn1 P89L heterozygote, and homozygote mutant mice. (WT n = 3, heterozygote n = 4, homozygote n = 3) Data are represented as means ± S.E.M. (*p<0.05, **p<0.01 Tukey-Kramer’s multiple comparisons test).

Fig 6. Comprehensive behavioral analysis of Nlgn1 P89L mice.

(A-C) Three-chamber social interaction test. (A) A stranger mouse was placed in one of the side chambers in a wired cage, and an empty wired cage was placed in the opposite chamber. Time spent in each chamber. (B) Time spent sniffing the stranger or the inanimate object. Both WT and Nlgn1 P89L (P/L) mice spent more time sniffing the stranger than the empty cage. (C) Social preference index was lower in Nlgn1 P89L (P/L) mice. Preference index = ((Time sniffing the stranger/ (Time sniffing the stranger + Time sniffing the inanimate)) x 100) - 50. *p<0.05, **p<0.01. Two-way repeated measures ANOVA, Bonferroni post-test (A, B). t-test (C), n = 15 for WT, n = 15 for heterozygous mutant. (D) Caged social interaction test in the open field. The test consists of two sessions, a 10 min habituation, followed by a 10-min test. Time spent around the cage during the habituation phase with an inanimate cage was identical between genotypes, however, time spent around the cage with an age-matched unfamiliar male mouse in the test session was significantly less in Nlgn1 P89L hetrozygote (P/L) mice compared to WT. **p<0.01. t-test, n = 15 for WT, n = 16 for heterozygous mutant. (E) Wins frequency in the social dominance tube test. Nlgn1 P89L (P/L) heterozygote mice had a lower winning rate. *p<0.05. Chi-square test. (F) The number of ultrasonic vocalizations emissions at postnatal day 7 induced by maternal-separation during 5 min session. (G) The number of ultrasonic vocalizations emissions from adult male mice during 10 min session. Adult female mouse was presented as stimuli. (H, I) Morris water maze to assess hippocampal-dependent spatial learning and memory. On day 1 and 2, mice were trained to find a visible platform in the water maze. On day 3 to 9, mice were trained to find a hidden platform. On day 10, spatial memory was assessed with the platform removed as a probe test (H) Quantification of latency to reach goal across days of training session from day 1 to 9. Two-way repeated measures ANOVA. (I) Quantification of time spent in each quadrant in a probe test session at day 10. TAR, ADJ and OPP indicates the target, adjacent, and opposite quadrant, respectively. WT showed significant preference for the target quadrant than the other quadrants, whereas Nlgn1 P89L (P/L) mice do not show the preference for the target quadrant compared to adjacent quadrants. **p<0.001, t-test. n = 15 for WT, n = 16 for heterozygous mutant. Data are represented as means ± S.E.M. See S3 Table for all statistics.